Research on the Volume Prediction of Sorting Center Based on ARIMA-Random Forest

In terms of sorting center volume forecasting, the traditional ARIMA model is difficult to cope with the rapid changes in the market and environment, so an innovative new method based on machine learning is proposed. Then features such as trend, period and lag are constructed for each sorting center to accurately capture the temporal features and improve the prediction accuracy. By comparing various models such as random forest, neural network, support vector machine and linear regression, and comprehensively considering the evaluation indexes such as MSE, MAE and RMSE, the random forest model with optimal performance is finally selected for cargo volume prediction. In the forecasting process, rolling forecasts are used to ensure that the model can be updated in real time and adapt to new market changes. In addition, considering the impact of changes in transport routes on cargo volume prediction, the model is further optimized by introducing transport network information to improve prediction accuracy. This comprehensiveness and flexibility make this method more applicable and effective in practical applications.